3. EMPIRICAL ANALYSIS
3.1. Micro-level analysis of six case studies
3.1.5. Case Study 3: Kosovo
After the Republic of Yugoslavia ceased to exist, Kosovo lost its autonomy. This was accompanied by large increases in levels of poverty and reductions in the living standards of some population groups. Ethnic, religious and social differences between Serbs and Albanians in particular led to a civil war in 1998. The war resulted in the death of over 13,000 people, large damage to housing and public infrastructure and the displacement of more than 1 million people
42 (UNHCR 2011). In 1999, NATO intervened in the region and the UN declared it as a Protectorate ever since.
In this section, we analyse the impact of the 1998-1999 Kosovo war on women‘s roles and activities, and their contribution to household welfare and community economic recovery. This empirical analysis is based on the 2000 Kosovo Living Standards Measurement Survey (KLSS), conducted just over a year later after the end of the NATO campaign. The sample includes 2,880 households and 17,917 individuals. The survey was designed to be nationally representative. It contains detailed information on individuals and households allowing us to analyse the short term impact of violence in terms of household composition, distribution of activities within households, participation in labour markets, and welfare indicators.
The survey also includes detailed self-reported information on the exposure of individuals and households to the 1998-1999 conflict in terms of displacement and the situation of current living conditions (whether the household dwelling was damaged or whether the household is living in temporary shelters due to the violence). We use this information to identify households exposed to conflict, assigning a value of one to those that reported either that they were living in damaged dwelling or in temporary shelters, and for those in which the entire household was displaced.
Throughout the analysis we will compare households directly affected by the incidence of violence with those not affected, disaggregating results by conflict incidence and gender. The sample involves 993 households living in regions not affected by conflict (35%), located in 134 communities, and 1,887 households in conflict affected regions during war (66%), located in 226 communities.
a) Impact of conflict on gender roles
Table 1 reports our main findings on the impact of the 1998-1999 conflict on household composition. In general, we find that households in Kosovo face low dependency ratios (below 1 dependant for each working age member). Female-headed households, almost entirely headed by a single woman, represent 8-9% of the overall sample. The results in table 1 show further that households in conflict-affected households are significantly likely to have less young children, more older children and less old age household members. Households are more likely to be headed by a woman in households less affected by the conflict.
The comparison of results between female- and male-headed households shows that dependency rates (the share of children and old age members in the total household size) are significantly lower in male-headed households affected by the conflict. Dependency rates are higher in female-headed households affected by the conflict, but the estimates are not statistically significant (possibly due to small cell sizes). Female-headed households that were most affected by the conflict are significantly more likely to be headed by single women and less likely to be headed by married women. This finding is in line with existing evidence discussed in chapter 2.
43 Table KS 1. Impact of conflict on household composition changes in Kosovo
All Female headed households Male headed households
Indicator Obs Non-
affected Conflict-
affected Diff t-test Obs Non-
affected Conflict-
affected Diff t-test Obs Non-
affected Conflict-
affected Diff t-test HH size (average) 2,880 6.064 6.304 -0.239 * 239 3.871 4.575 -0.704 * 2,641 6.291 6.449 -0.157 n.s.
Composition
Share of children 0-6 yrs 2,880 0.133 0.113 0.020 *** 239 0.064 0.089 -0.025 n.s. 2,641 0.140 0.115 0.025 ***
Share of children 7-15 yrs 2,880 0.142 0.184 -0.042 *** 239 0.139 0.146 -0.007 n.s. 2,641 0.142 0.187 -0.045 ***
Share of males 16-65 yrs 2,880 0.302 0.306 -0.004 n.s. 239 0.152 0.198 -0.046 * 2,641 0.318 0.316 0.002 n.s.
Share of females 16-65 yrs 2,880 0.321 0.319 0.001 n.s. 239 0.409 0.421 -0.011 n.s. 2,641 0.311 0.311 0.001 n.s.
Share of elders >65 yrs 2,880 0.102 0.078 0.025 *** 239 0.236 0.146 0.090 ** 2,641 0.089 0.072 0.016 **
Dependency ratio (<16
and >65 / 16 to 65 yrs) 2,783 0.769 0.797 -0.028 n.s. 209 0.755 0.801 -0.046 n.s. 2,574 0.770 0.796 -0.027 n.s.
Female/male ratio
(16 to 65 yrs) 10,772 0.512 0.514 -0.002 n.s. 618 0.658 0.622 0.036 n.s. 10,154 0.502 0.507 -0.005 n.s.
Type of households
Single female headed
household 2,880 0.091 0.077 0.013 n.s. 239 0.968 1.000 -0.032 ** 2,641 0.000 0.000 0.000 n.a.
Married female headed
households 2,880 0.003 0.000 0.003 ** 239 0.032 0.000 0.032 ** 2,641 0.000 0.000 0.000 n.a.
Single male headed
household 2,880 0.057 0.067 -0.010 n.s. 239 0.000 0.000 0.000 n.a. 2,641 0.063 0.073 -0.010 n.s.
Married male headed
households 2,880 0.849 0.855 -0.006 n.s. 239 0.000 0.000 0.000 n.a. 2,641 0.937 0.927 0.010 n.s.
HHs directly affected by conflict
Displacement (%) 2,880 0.000 0.722 -0.722 n.a. 239 0.000 0.589 -0.589 n.a. 2,641 0.000 0.733 -0.733 n.a.
Damage to dwelling (%) 2,865 0.000 0.592 -0.592 n.a. 238 0.000 0.603 -0.603 n.a. 2,627 0.000 0.591 -0.591 n.a.
Other household characteristics
Access to land
(% with >0.5 ha) 2,880 0.403 0.568 -0.165 *** 239 0.226 0.459 -0.233 *** 2,641 0.421 0.577 -0.156 ***
Rural location
(% of households) 2,880 0.550 0.749 -0.199 *** 239 0.462 0.664 -0.202 *** 2,641 0.559 0.756 -0.197 ***
Source: Own calculations based on KLSS 2000.
Notes: *** statistically significant at 1%, ** statistically significant at 5%, * statistically significant at 10%. n.s. indicates results that are not statistically significant.
44 Table 2 provides comparative estimates on female and male participation in labour markets in households affected and not affected by the conflict. The results show significant effects of the Kosovo conflict on the economic roles of women and men. Similarly to other case studies, we observe that female household heads affected by the war are 9.3% more likely to be employed than those not affected by the conflict. Other female members in households headed by a woman and affected by war are 3% more likely to be employed than women living in households not affected by violence. This result is statistically significant at the 1% level of significance. In the case of men, we found more mixed results: male household heads are more likely to be employed if they were directly affected by conflict (by 5.8%). However, other men living in households affected by the conflict (and that are not household heads) are less likely to be employed by the same proportion. Men in households affected by the conflict are also significantly more likely to be inactive. These results confirm the hypothesis that women tend to join the labour market in situations of conflict through an ‗added worker‘ effect (hypothesis 2).
In line with some findings in the literature reviewed in chapter 2, men in Kosovo are more likely to be unemployed when their household was directly exposed to violence (in this case, with the exception of men that are household heads).
Table 3 presents more disaggregated information on the characteristics of females and males employed in Kosovo. We find that both men and women affected by conflict work similar hours than their counterparts not directly affected by the conflict. We do find that women work fewer hours per week than men (9 hours per week less). We cannot assess differences in earnings given that the information on incomes from the KLSS 2000 is incomplete (Douarin et al. 2010). We find further that women affected by conflict are less likely to be working as waged employees, and considerably more likely to be engaged in self-employment, both in agriculture and other sectors, when compared to women not affected by conflict. Although men affected by conflict exhibit similar patterns, their access to wage employment is higher than that of women. Men also depend much less on self-employment agricultural work in comparison to women. This finding is in line with accounts in the literature and other empirical evidence we obtained for other case studies that women that join labour markets in conflict contexts tend to be employed in low-paid, low-skilled jobs.
b) Impact of changes in gender roles on household welfare
We have analysed the impact of changes in the economic roles of men and women as a result of the ethnic conflict in Kosovo on household welfare. Household welfare is again measured in terms of household consumption expenditure per capita. The results in table 4 show that households affected by the conflict have lower values of consumption expenditure per capita, although in general the differences are not statistically significant. Further econometric analysis will be needed in order to better understand these results, controlling or other factors affecting the levels of consumption across households. Contrary to some of the other case studies, we have found that increased levels of employment against women in Kosovo has resulted in statistically significant negative effects on household consumption expenditure in households directly affected by the conflict. This effect is particularly pronounced among female-headed households, and explained by the adoption of low status jobs by women in Kosovo (see table 3).
45 Table KS 2. Impact of conflict on economic roles in Kosovo
Females Males
Position in HH / Indicator Obs Non-
affected Conflict-
affected Diff t-
test Obs Non-
affected Conflict-
affected Diff t- test
Labor Force Participation, HH Heads
Employed 239 0.215 0.308 -0.093 n.s. 2,641 0.590 0.648 -0.058 ***
Unemployed 239 0.043 0.014 0.029 n.s. 2,641 0.079 0.079 0.000 n.s.
Inactive 239 0.742 0.678 0.064 n.s. 2,641 0.331 0.272 0.059 ***
Labor Force Participation, Others
Employed 5,888 0.222 0.252 -0.030 *** 3,074 0.512 0.454 0.058 ***
Unemployed 5,888 0.064 0.057 0.007 n.s. 3,074 0.167 0.149 0.018 n.s.
Inactive 5,888 0.714 0.691 0.023 * 3,074 0.321 0.397 -0.076 ***
Source: Own calculations based on KLSS 2000.
Notes: *** statistically significant at 1%, ** statistically significant at 5%, * statistically significant at 10%. n.s. indicates results that are not statistically significant.
Table KS 3. Impact of conflict on employment characteristics in Kosovo
Females Males
Indicator Obs Non-
affected Conflict-
affected Diff t-
test Obs Non-
affected Conflict-
affected Diff t- test Average hours worked per week 1,442 33.599 32.809 0.791 n.s. 3,041 42.705 41.803 0.902 n.s.
Type of employment
Wage employment 1,443 0.408 0.230 0.178 *** 3,041 0.524 0.503 0.021 n.s.
Self-employment - agriculture 1,443 0.389 0.523 -0.134 *** 3,041 0.250 0.318 -0.068 ***
Self-employment – not
agriculture 1,443 0.203 0.247 -0.044 * 3,041 0.226 0.179 0.047 ***
Access to secondary jobs 1,490 0.013 0.017 -0.004 n.s. 3,118 0.056 0.080 -0.024 **
Source: Own calculations based on KLSS 2000.
Notes: *** statistically significant at 1%, ** statistically significant at 5%, * statistically significant at 10%. n.s. indicates results that are not statistically significant.
(1) Excluding zeros for hours worked.
46 Table KS 4. Impact of changes in gender roles on household welfare in Kosovo
All Female headed households Male headed households
Indicator / Status of HH Head Obs Non-
affected Conflict-
affected Diff t-test Obs Non-
affected Conflict-
affected Diff t-test Obs Non-
affected Conflict-
affected Diff t-test
Average expenditure per capita
All households
No female member employed 1,657 115 112 2 n.s. 150 98 116 -19 * 1,507 117 112 5 n.s.
At least 1 female member employed 1,184 138 130 9 * 89 152 118 33 * 1,095 137 130 7 n.s.
When hh head is employed
No female member employed 824 128 121 7 n.s. n.a. n.a. n.a. n.a. n.a. 824 128 121 7 n.s.
At least 1 female member employed 879 142 134 9 n.s. 65 146 106 40 ** 814 142 136 6 n.s.
When hh head is not employed
No female member employed 833 104 102 2 n.s. 150 98 116 -19 * 683 105 99 6 n.s.
At least 1 female member employed 305 128 118 10 n.s. 24 164 156 8 n.s. 281 124 115 10 n.s.
Source: Own calculations based on KLSS 2000.
Notes: *** statistically significant at 1%, ** statistically significant at 5%, * statistically significant at 10%. n.s. indicates results that are not statistically significant. Consumption expenditure is provided in euro.
47 c) Impact of changes in gender roles on community welfare
In table 5, we report differences in average welfare at the community level associated to changes in gender roles due to conflict. For this purpose, we have estimated aggregate average consumption per capita in each community. In order to assess further the impact of women‘s involvement in economic roles due to conflict, we have identified communities according to the share of women in employment (below or above average) (table 5). In general, the differences are not significant (most likely due to small cell sizes). We observe nonetheless that communities that have above the average shares of female workers have higher levels of consumption per capita (3% increase). Table 5 shows an opposite result for communities with below the average shares of female employment (5% reduction in consumption welfare levels).
Table KS 5. Impact of changes in gender roles on community welfare in Kosovo
All
Notes: *** statistically significant at 1%, ** statistically significant at 5%, * statistically significant at 10%.
n.s. indicates results that are not statistically significant. Consumption expenditure is provided in euro.
This result seems to suggest that although rises in female employment in Kosovo due to the war are associated with reductions in welfare at the household level (and thus not confirming hypothesis 5), higher rates of female employment have positive benefits for community welfare levels in areas affected by the war in Kosovo (thus confirming hypothesis 6). This is an important result but further confirmation is needed through the use of more sophisticated econometric techniques outside the scope of this phase of the study. We have run simple regression analysis (as in other case studies) that still shows statistically insignificant results. The results in those regressions are not, however, reliable. In the case of Kosovo, the conflict exposure variable is provided at the household-level and is self-reported. It is therefore very likely that the results are contaminated by serious endogeneity problems (resulting from potential omitted variable and reverse causality biases). These issues can only be addressed through the use of an instrumental variable approach, to be pursued in phase II of the study.